Parametric fusion of complex landmark observations present within the road network by utilizing bundle-adjustment-based Full-SLAM

摘要: Maps representing the detailed features of road network are becoming more and important for autonomous driving next generation driver assistance systems. The mapping by specially equipped vehicles through well-known map providers leads to usually quarterly updates, which might result in problems encountered case that information is outdated. Furthermore, provided maps could lack details relevant As an alternative, data can be acquired common subsequently fused at backend-side detailed, up-to-date maps. deduction consisting point-shaped landmarks utilizing approaches SLAM not trivial but well studied. However, comprised many also non-point-shaped landmarks, such as crossroads, roundabouts, sign gantries, traffic islands crosswalks. reduction these complex ones or independent consideration their basic features, points lines, involves a severe loss useful and/or accuracy is, therefore, appreciated. Instead, we propose novel approach parametric description directed acyclic graph handle landmark observations extended formulation Bundle-Adjustment-based Full-SLAM. We focus on challenges imposed fusion attribute interdependencies, dynamic counts, partial attributes due occlusions. show convergence consistency our hybrid, real-world scenario roundabout observations. presented way allows straightforward adaptation other kinds defining appropriate description.